Intern, ML Compiler (Kernel Auto-Tuning)
Samsung Semiconductor2 days ago
San Jose, CA, USAIntern
Responsibilities
- Design search spaces for Triton kernel configurations, including tile sizes, launch-grid mapping, on-chip buffer budgets, and load style.
- Build legality rules that eliminate configurations that cannot compile or fit in on-chip memory.
- Develop a kernel cost model calibrated against simulators and profilers to rank tuning candidates.
- Build a benchmarking harness and persistent tuning cache.
- Integrate the tuner with the torch.compile Inductor autotuning path and Triton’s autotuner.
- Extend tuning to tensor data placement and export tuned costs for use by a placement pass.
- Measure kernel-level and end-to-end LLM inference performance gains.
- Collaborate with PyTorch and Triton developers on upstream autotuning hooks and deliver documented, tested code.
- Present results in weekly syncs and a final presentation.
Requirements
- Pursuing a master’s or PhD in Computer Science, Computer Engineering, or Electrical Engineering, with at least one academic quarter or semester remaining.
- Strong Python and C++ skills.
- Experience writing or tuning Triton kernels or working with PyTorch compiler internals, including torch.compile, FX, or Inductor.
- Knowledge of compilers or performance engineering, including IR and passes, cost models, roofline analysis, and profiling.
Benefits
- Hourly base pay of $28—$68 USD.
- Daily onsite work at Samsung’s San Jose Headquarters under a flexible work policy.
- Relocation stipend for candidates living 50 or more miles from the new office.
- Monthly housing stipend for a new place during the summer internship.
- Medical, dental, and vision benefits.
- Charitable giving match, emotional wellness apps and paid therapy sessions, onsite café and gym, virtual fitness classes, and a flexible work environment.